Spatial multivariate regressions with panel data

We develop a new Bayesian estimator that is able to deal with multivariate panel data structure in the presence of spatial correlation. The analysis of panel data introduced here allows us to analyze not only the fixed effect but also the random effect model. This work extends the previous study undertaken by Gamerman and Moreira (2004) which only spatial scale is considered. To estimate the random effect model we use the hierarchical analysis that can be applied to estimate some categories of longitudinal data models. The Monte Carlo simulations demonstrate the ability of this new estimator to replicate quite well simulated data. To show the empirical relevance of this new estimator we apply it to the deforestation data in the Brazilian Amazon.